Research map / Continual learning / Domain and prompt-based
S-Prompts Learning with Pre-trained Transformers: An Occam’s Razor for Domain Incremental Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use S-liPrompts (language-image prompts) instead of S-iPrompts | S-Prompts · CDDB-Hard | average accuracy: 74.51 -> 88.65 | Helped |
| S-liPrompts vs best exemplar-free baseline L2P | exemplar-free DIL · DomainNet | average accuracy: 40.15 -> 67.78 | Helped |
| S-liPrompts vs joint-training upper bound | S-liPrompts · DomainNet | average accuracy: 64.08 -> 67.78 | Helped |
| use growing classifier pool (replace fixed FC length) (ablation) | S-iPrompts · CDDB-Hard | average accuracy: 52.07 -> 74.51 | Helped |
| use independent prompts per domain (replace fixed FC weights/shared) (ablation) | S-iPrompts · CDDB-Hard | average accuracy: 72.13 -> 74.51 | Helped |
| use growing language prompt length (replace fixed) (ablation) | S-liPrompts · CDDB-Hard | average accuracy: 52.80 -> 88.65 | Helped |
| use independent language prompt weights (replace fixed weights) (ablation) | S-liPrompts · CDDB-Hard | average accuracy: 72.29 -> 88.65 | Helped |
| use K-NN domain selection instead of random domain selection (ablation) | S-liPrompts · CDDB-Hard | average accuracy: 80.12 -> 88.65 | Helped |
| vary K-Means K from 1 to 9 (ablation) | S-liPrompts · CDDB-Hard | average accuracy: 88.72 -> 88.32 | No clear effect |
| train on all domains instead of first domain only (zero-shot) (ablation) | S-liPrompts · CDDB-Hard | average accuracy: 66.57 -> 88.65 | Helped |
| too small image prompt length (ablation) | S-liPrompts · CDDB-Hard | average accuracy | Hurt |
Closest papers by meaning
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.91)
- Mixture of Experts Meets Prompt-Based Continual Learning (NeurIPS 2024 · similarity 0.89)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.89)
- SEC-Prompt:SEmantic Complementary Prompting for Few-Shot Class-Incremental Learning (CVPR 2025 · similarity 0.89)
- Visual Prompt Tuning in Null Space for Continual Learning (NeurIPS 2024 · similarity 0.88)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.